Same Pre-training Loss, Better Downstream: Implicit Bias Matters for Language Models.
Hong LiuSang Michael XieZhiyuan LiTengyu MaPublished in: CoRR (2022)
Keyphrases
- language model
- language modeling
- speech recognition
- n gram
- probabilistic model
- information retrieval
- document retrieval
- test collection
- retrieval model
- language modelling
- statistical language models
- query terms
- context sensitive
- ad hoc information retrieval
- vector space model
- smoothing methods
- language models for information retrieval
- relevance model
- okapi bm
- document ranking
- translation model
- query specific
- query expansion
- statistical language modeling
- language model for information retrieval
- passage retrieval
- machine learning
- training data
- training set
- text classification